]>
Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/exec.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | /* | |
9 | * #!-checking implemented by tytso. | |
10 | */ | |
11 | /* | |
12 | * Demand-loading implemented 01.12.91 - no need to read anything but | |
13 | * the header into memory. The inode of the executable is put into | |
14 | * "current->executable", and page faults do the actual loading. Clean. | |
15 | * | |
16 | * Once more I can proudly say that linux stood up to being changed: it | |
17 | * was less than 2 hours work to get demand-loading completely implemented. | |
18 | * | |
19 | * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead, | |
20 | * current->executable is only used by the procfs. This allows a dispatch | |
21 | * table to check for several different types of binary formats. We keep | |
22 | * trying until we recognize the file or we run out of supported binary | |
613cc2b6 | 23 | * formats. |
1da177e4 LT |
24 | */ |
25 | ||
b89999d0 | 26 | #include <linux/kernel_read_file.h> |
1da177e4 LT |
27 | #include <linux/slab.h> |
28 | #include <linux/file.h> | |
9f3acc31 | 29 | #include <linux/fdtable.h> |
ba92a43d | 30 | #include <linux/mm.h> |
615d6e87 | 31 | #include <linux/vmacache.h> |
1da177e4 LT |
32 | #include <linux/stat.h> |
33 | #include <linux/fcntl.h> | |
ba92a43d | 34 | #include <linux/swap.h> |
74aadce9 | 35 | #include <linux/string.h> |
1da177e4 | 36 | #include <linux/init.h> |
6e84f315 | 37 | #include <linux/sched/mm.h> |
f7ccbae4 | 38 | #include <linux/sched/coredump.h> |
3f07c014 | 39 | #include <linux/sched/signal.h> |
6a3827d7 | 40 | #include <linux/sched/numa_balancing.h> |
29930025 | 41 | #include <linux/sched/task.h> |
ca5b172b | 42 | #include <linux/pagemap.h> |
cdd6c482 | 43 | #include <linux/perf_event.h> |
1da177e4 LT |
44 | #include <linux/highmem.h> |
45 | #include <linux/spinlock.h> | |
46 | #include <linux/key.h> | |
47 | #include <linux/personality.h> | |
48 | #include <linux/binfmts.h> | |
1da177e4 | 49 | #include <linux/utsname.h> |
84d73786 | 50 | #include <linux/pid_namespace.h> |
1da177e4 LT |
51 | #include <linux/module.h> |
52 | #include <linux/namei.h> | |
1da177e4 LT |
53 | #include <linux/mount.h> |
54 | #include <linux/security.h> | |
55 | #include <linux/syscalls.h> | |
8f0ab514 | 56 | #include <linux/tsacct_kern.h> |
9f46080c | 57 | #include <linux/cn_proc.h> |
473ae30b | 58 | #include <linux/audit.h> |
6341c393 | 59 | #include <linux/tracehook.h> |
5f4123be | 60 | #include <linux/kmod.h> |
6110e3ab | 61 | #include <linux/fsnotify.h> |
5ad4e53b | 62 | #include <linux/fs_struct.h> |
3d5992d2 | 63 | #include <linux/oom.h> |
0e028465 | 64 | #include <linux/compat.h> |
b44a7dfc | 65 | #include <linux/vmalloc.h> |
0f212204 | 66 | #include <linux/io_uring.h> |
1446e1df | 67 | #include <linux/syscall_user_dispatch.h> |
1da177e4 | 68 | |
7c0f6ba6 | 69 | #include <linux/uaccess.h> |
1da177e4 | 70 | #include <asm/mmu_context.h> |
b6a2fea3 | 71 | #include <asm/tlb.h> |
43d2b113 KH |
72 | |
73 | #include <trace/events/task.h> | |
a6f76f23 | 74 | #include "internal.h" |
1da177e4 | 75 | |
4ff16c25 DS |
76 | #include <trace/events/sched.h> |
77 | ||
56305aa9 EB |
78 | static int bprm_creds_from_file(struct linux_binprm *bprm); |
79 | ||
d6e71144 AC |
80 | int suid_dumpable = 0; |
81 | ||
e4dc1b14 | 82 | static LIST_HEAD(formats); |
1da177e4 LT |
83 | static DEFINE_RWLOCK(binfmt_lock); |
84 | ||
8fc3dc5a | 85 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 86 | { |
1da177e4 | 87 | write_lock(&binfmt_lock); |
74641f58 IK |
88 | insert ? list_add(&fmt->lh, &formats) : |
89 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 90 | write_unlock(&binfmt_lock); |
1da177e4 LT |
91 | } |
92 | ||
74641f58 | 93 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 94 | |
f6b450d4 | 95 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 96 | { |
1da177e4 | 97 | write_lock(&binfmt_lock); |
e4dc1b14 | 98 | list_del(&fmt->lh); |
1da177e4 | 99 | write_unlock(&binfmt_lock); |
1da177e4 LT |
100 | } |
101 | ||
102 | EXPORT_SYMBOL(unregister_binfmt); | |
103 | ||
104 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
105 | { | |
106 | module_put(fmt->module); | |
107 | } | |
108 | ||
90f8572b EB |
109 | bool path_noexec(const struct path *path) |
110 | { | |
111 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
112 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
113 | } | |
7fea21ae | 114 | EXPORT_SYMBOL_GPL(path_noexec); |
90f8572b | 115 | |
69369a70 | 116 | #ifdef CONFIG_USELIB |
1da177e4 LT |
117 | /* |
118 | * Note that a shared library must be both readable and executable due to | |
119 | * security reasons. | |
120 | * | |
121 | * Also note that we take the address to load from from the file itself. | |
122 | */ | |
1e7bfb21 | 123 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 124 | { |
72c2d531 | 125 | struct linux_binfmt *fmt; |
964bd183 | 126 | struct file *file; |
91a27b2a | 127 | struct filename *tmp = getname(library); |
964bd183 | 128 | int error = PTR_ERR(tmp); |
47c805dc AV |
129 | static const struct open_flags uselib_flags = { |
130 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
62fb4a15 | 131 | .acc_mode = MAY_READ | MAY_EXEC, |
f9652e10 AV |
132 | .intent = LOOKUP_OPEN, |
133 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 134 | }; |
964bd183 | 135 | |
6e8341a1 AV |
136 | if (IS_ERR(tmp)) |
137 | goto out; | |
138 | ||
f9652e10 | 139 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
140 | putname(tmp); |
141 | error = PTR_ERR(file); | |
142 | if (IS_ERR(file)) | |
1da177e4 LT |
143 | goto out; |
144 | ||
633fb6ac KC |
145 | /* |
146 | * may_open() has already checked for this, so it should be | |
147 | * impossible to trip now. But we need to be extra cautious | |
148 | * and check again at the very end too. | |
149 | */ | |
db19c91c | 150 | error = -EACCES; |
0fd338b2 KC |
151 | if (WARN_ON_ONCE(!S_ISREG(file_inode(file)->i_mode) || |
152 | path_noexec(&file->f_path))) | |
1da177e4 LT |
153 | goto exit; |
154 | ||
2a12a9d7 | 155 | fsnotify_open(file); |
6110e3ab | 156 | |
1da177e4 | 157 | error = -ENOEXEC; |
1da177e4 | 158 | |
72c2d531 AV |
159 | read_lock(&binfmt_lock); |
160 | list_for_each_entry(fmt, &formats, lh) { | |
161 | if (!fmt->load_shlib) | |
162 | continue; | |
163 | if (!try_module_get(fmt->module)) | |
164 | continue; | |
1da177e4 | 165 | read_unlock(&binfmt_lock); |
72c2d531 AV |
166 | error = fmt->load_shlib(file); |
167 | read_lock(&binfmt_lock); | |
168 | put_binfmt(fmt); | |
169 | if (error != -ENOEXEC) | |
170 | break; | |
1da177e4 | 171 | } |
72c2d531 | 172 | read_unlock(&binfmt_lock); |
6e8341a1 | 173 | exit: |
1da177e4 LT |
174 | fput(file); |
175 | out: | |
176 | return error; | |
1da177e4 | 177 | } |
69369a70 | 178 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 179 | |
b6a2fea3 | 180 | #ifdef CONFIG_MMU |
ae6b585e ON |
181 | /* |
182 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
183 | * use a lot of memory, account these pages in current->mm temporary | |
184 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
185 | * change the counter back via acct_arg_size(0). | |
186 | */ | |
0e028465 | 187 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
188 | { |
189 | struct mm_struct *mm = current->mm; | |
190 | long diff = (long)(pages - bprm->vma_pages); | |
191 | ||
192 | if (!mm || !diff) | |
193 | return; | |
194 | ||
195 | bprm->vma_pages = pages; | |
3c77f845 | 196 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
197 | } |
198 | ||
0e028465 | 199 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
200 | int write) |
201 | { | |
202 | struct page *page; | |
203 | int ret; | |
9beae1ea | 204 | unsigned int gup_flags = FOLL_FORCE; |
b6a2fea3 OW |
205 | |
206 | #ifdef CONFIG_STACK_GROWSUP | |
207 | if (write) { | |
d05f3169 | 208 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
209 | if (ret < 0) |
210 | return NULL; | |
211 | } | |
212 | #endif | |
9beae1ea LS |
213 | |
214 | if (write) | |
215 | gup_flags |= FOLL_WRITE; | |
216 | ||
1e987790 DH |
217 | /* |
218 | * We are doing an exec(). 'current' is the process | |
219 | * doing the exec and bprm->mm is the new process's mm. | |
220 | */ | |
5b78ed24 | 221 | mmap_read_lock(bprm->mm); |
64019a2e | 222 | ret = get_user_pages_remote(bprm->mm, pos, 1, gup_flags, |
5b56d49f | 223 | &page, NULL, NULL); |
5b78ed24 | 224 | mmap_read_unlock(bprm->mm); |
b6a2fea3 OW |
225 | if (ret <= 0) |
226 | return NULL; | |
227 | ||
655c16a8 ON |
228 | if (write) |
229 | acct_arg_size(bprm, vma_pages(bprm->vma)); | |
b6a2fea3 OW |
230 | |
231 | return page; | |
232 | } | |
233 | ||
234 | static void put_arg_page(struct page *page) | |
235 | { | |
236 | put_page(page); | |
237 | } | |
238 | ||
b6a2fea3 OW |
239 | static void free_arg_pages(struct linux_binprm *bprm) |
240 | { | |
241 | } | |
242 | ||
243 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
244 | struct page *page) | |
245 | { | |
246 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
247 | } | |
248 | ||
249 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
250 | { | |
eaccbfa5 | 251 | int err; |
b6a2fea3 OW |
252 | struct vm_area_struct *vma = NULL; |
253 | struct mm_struct *mm = bprm->mm; | |
254 | ||
490fc053 | 255 | bprm->vma = vma = vm_area_alloc(mm); |
b6a2fea3 | 256 | if (!vma) |
eaccbfa5 | 257 | return -ENOMEM; |
bfd40eaf | 258 | vma_set_anonymous(vma); |
b6a2fea3 | 259 | |
d8ed45c5 | 260 | if (mmap_write_lock_killable(mm)) { |
f268dfe9 MH |
261 | err = -EINTR; |
262 | goto err_free; | |
263 | } | |
b6a2fea3 OW |
264 | |
265 | /* | |
266 | * Place the stack at the largest stack address the architecture | |
267 | * supports. Later, we'll move this to an appropriate place. We don't | |
268 | * use STACK_TOP because that can depend on attributes which aren't | |
269 | * configured yet. | |
270 | */ | |
aacb3d17 | 271 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
272 | vma->vm_end = STACK_TOP_MAX; |
273 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 274 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 275 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
462e635e | 276 | |
b6a2fea3 | 277 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 278 | if (err) |
b6a2fea3 | 279 | goto err; |
b6a2fea3 OW |
280 | |
281 | mm->stack_vm = mm->total_vm = 1; | |
d8ed45c5 | 282 | mmap_write_unlock(mm); |
b6a2fea3 | 283 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 284 | return 0; |
b6a2fea3 | 285 | err: |
d8ed45c5 | 286 | mmap_write_unlock(mm); |
f268dfe9 | 287 | err_free: |
eaccbfa5 | 288 | bprm->vma = NULL; |
3928d4f5 | 289 | vm_area_free(vma); |
b6a2fea3 OW |
290 | return err; |
291 | } | |
292 | ||
293 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
294 | { | |
295 | return len <= MAX_ARG_STRLEN; | |
296 | } | |
297 | ||
298 | #else | |
299 | ||
0e028465 | 300 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
301 | { |
302 | } | |
303 | ||
0e028465 | 304 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
305 | int write) |
306 | { | |
307 | struct page *page; | |
308 | ||
309 | page = bprm->page[pos / PAGE_SIZE]; | |
310 | if (!page && write) { | |
311 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
312 | if (!page) | |
313 | return NULL; | |
314 | bprm->page[pos / PAGE_SIZE] = page; | |
315 | } | |
316 | ||
317 | return page; | |
318 | } | |
319 | ||
320 | static void put_arg_page(struct page *page) | |
321 | { | |
322 | } | |
323 | ||
324 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
325 | { | |
326 | if (bprm->page[i]) { | |
327 | __free_page(bprm->page[i]); | |
328 | bprm->page[i] = NULL; | |
329 | } | |
330 | } | |
331 | ||
332 | static void free_arg_pages(struct linux_binprm *bprm) | |
333 | { | |
334 | int i; | |
335 | ||
336 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
337 | free_arg_page(bprm, i); | |
338 | } | |
339 | ||
340 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
341 | struct page *page) | |
342 | { | |
343 | } | |
344 | ||
345 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
346 | { | |
347 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
348 | return 0; | |
349 | } | |
350 | ||
351 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
352 | { | |
353 | return len <= bprm->p; | |
354 | } | |
355 | ||
356 | #endif /* CONFIG_MMU */ | |
357 | ||
358 | /* | |
359 | * Create a new mm_struct and populate it with a temporary stack | |
360 | * vm_area_struct. We don't have enough context at this point to set the stack | |
361 | * flags, permissions, and offset, so we use temporary values. We'll update | |
362 | * them later in setup_arg_pages(). | |
363 | */ | |
9cc64cea | 364 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
365 | { |
366 | int err; | |
367 | struct mm_struct *mm = NULL; | |
368 | ||
369 | bprm->mm = mm = mm_alloc(); | |
370 | err = -ENOMEM; | |
371 | if (!mm) | |
372 | goto err; | |
373 | ||
c31dbb14 KC |
374 | /* Save current stack limit for all calculations made during exec. */ |
375 | task_lock(current->group_leader); | |
376 | bprm->rlim_stack = current->signal->rlim[RLIMIT_STACK]; | |
377 | task_unlock(current->group_leader); | |
378 | ||
b6a2fea3 OW |
379 | err = __bprm_mm_init(bprm); |
380 | if (err) | |
381 | goto err; | |
382 | ||
383 | return 0; | |
384 | ||
385 | err: | |
386 | if (mm) { | |
387 | bprm->mm = NULL; | |
388 | mmdrop(mm); | |
389 | } | |
390 | ||
391 | return err; | |
392 | } | |
393 | ||
ba2d0162 | 394 | struct user_arg_ptr { |
0e028465 ON |
395 | #ifdef CONFIG_COMPAT |
396 | bool is_compat; | |
397 | #endif | |
398 | union { | |
399 | const char __user *const __user *native; | |
400 | #ifdef CONFIG_COMPAT | |
38b983b3 | 401 | const compat_uptr_t __user *compat; |
0e028465 ON |
402 | #endif |
403 | } ptr; | |
ba2d0162 ON |
404 | }; |
405 | ||
406 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 407 | { |
0e028465 ON |
408 | const char __user *native; |
409 | ||
410 | #ifdef CONFIG_COMPAT | |
411 | if (unlikely(argv.is_compat)) { | |
412 | compat_uptr_t compat; | |
413 | ||
414 | if (get_user(compat, argv.ptr.compat + nr)) | |
415 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 416 | |
0e028465 ON |
417 | return compat_ptr(compat); |
418 | } | |
419 | #endif | |
420 | ||
421 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
422 | return ERR_PTR(-EFAULT); |
423 | ||
0e028465 | 424 | return native; |
1d1dbf81 ON |
425 | } |
426 | ||
1da177e4 LT |
427 | /* |
428 | * count() counts the number of strings in array ARGV. | |
429 | */ | |
ba2d0162 | 430 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
431 | { |
432 | int i = 0; | |
433 | ||
0e028465 | 434 | if (argv.ptr.native != NULL) { |
1da177e4 | 435 | for (;;) { |
1d1dbf81 | 436 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 437 | |
1da177e4 LT |
438 | if (!p) |
439 | break; | |
1d1dbf81 ON |
440 | |
441 | if (IS_ERR(p)) | |
442 | return -EFAULT; | |
443 | ||
6d92d4f6 | 444 | if (i >= max) |
1da177e4 | 445 | return -E2BIG; |
6d92d4f6 | 446 | ++i; |
9aea5a65 RM |
447 | |
448 | if (fatal_signal_pending(current)) | |
449 | return -ERESTARTNOHAND; | |
1da177e4 LT |
450 | cond_resched(); |
451 | } | |
452 | } | |
453 | return i; | |
454 | } | |
455 | ||
be619f7f EB |
456 | static int count_strings_kernel(const char *const *argv) |
457 | { | |
458 | int i; | |
459 | ||
460 | if (!argv) | |
461 | return 0; | |
462 | ||
463 | for (i = 0; argv[i]; ++i) { | |
464 | if (i >= MAX_ARG_STRINGS) | |
465 | return -E2BIG; | |
466 | if (fatal_signal_pending(current)) | |
467 | return -ERESTARTNOHAND; | |
468 | cond_resched(); | |
469 | } | |
470 | return i; | |
471 | } | |
472 | ||
d8b9cd54 | 473 | static int bprm_stack_limits(struct linux_binprm *bprm) |
655c16a8 ON |
474 | { |
475 | unsigned long limit, ptr_size; | |
476 | ||
655c16a8 ON |
477 | /* |
478 | * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM | |
479 | * (whichever is smaller) for the argv+env strings. | |
480 | * This ensures that: | |
481 | * - the remaining binfmt code will not run out of stack space, | |
482 | * - the program will have a reasonable amount of stack left | |
483 | * to work from. | |
484 | */ | |
485 | limit = _STK_LIM / 4 * 3; | |
486 | limit = min(limit, bprm->rlim_stack.rlim_cur / 4); | |
487 | /* | |
488 | * We've historically supported up to 32 pages (ARG_MAX) | |
489 | * of argument strings even with small stacks | |
490 | */ | |
491 | limit = max_t(unsigned long, limit, ARG_MAX); | |
492 | /* | |
493 | * We must account for the size of all the argv and envp pointers to | |
494 | * the argv and envp strings, since they will also take up space in | |
495 | * the stack. They aren't stored until much later when we can't | |
496 | * signal to the parent that the child has run out of stack space. | |
497 | * Instead, calculate it here so it's possible to fail gracefully. | |
c07dc0fe KC |
498 | * |
499 | * In the case of argc = 0, make sure there is space for adding a | |
500 | * empty string (which will bump argc to 1), to ensure confused | |
501 | * userspace programs don't start processing from argv[1], thinking | |
502 | * argc can never be 0, to keep them from walking envp by accident. | |
503 | * See do_execveat_common(). | |
655c16a8 | 504 | */ |
c07dc0fe | 505 | ptr_size = (max(bprm->argc, 1) + bprm->envc) * sizeof(void *); |
655c16a8 ON |
506 | if (limit <= ptr_size) |
507 | return -E2BIG; | |
508 | limit -= ptr_size; | |
509 | ||
510 | bprm->argmin = bprm->p - limit; | |
511 | return 0; | |
512 | } | |
513 | ||
1da177e4 | 514 | /* |
b6a2fea3 OW |
515 | * 'copy_strings()' copies argument/environment strings from the old |
516 | * processes's memory to the new process's stack. The call to get_user_pages() | |
517 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 518 | */ |
ba2d0162 | 519 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 520 | struct linux_binprm *bprm) |
1da177e4 LT |
521 | { |
522 | struct page *kmapped_page = NULL; | |
523 | char *kaddr = NULL; | |
b6a2fea3 | 524 | unsigned long kpos = 0; |
1da177e4 LT |
525 | int ret; |
526 | ||
527 | while (argc-- > 0) { | |
d7627467 | 528 | const char __user *str; |
1da177e4 LT |
529 | int len; |
530 | unsigned long pos; | |
531 | ||
1d1dbf81 ON |
532 | ret = -EFAULT; |
533 | str = get_user_arg_ptr(argv, argc); | |
534 | if (IS_ERR(str)) | |
1da177e4 | 535 | goto out; |
1da177e4 | 536 | |
1d1dbf81 ON |
537 | len = strnlen_user(str, MAX_ARG_STRLEN); |
538 | if (!len) | |
539 | goto out; | |
540 | ||
541 | ret = -E2BIG; | |
542 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 543 | goto out; |
1da177e4 | 544 | |
b6a2fea3 | 545 | /* We're going to work our way backwords. */ |
1da177e4 | 546 | pos = bprm->p; |
b6a2fea3 OW |
547 | str += len; |
548 | bprm->p -= len; | |
655c16a8 ON |
549 | #ifdef CONFIG_MMU |
550 | if (bprm->p < bprm->argmin) | |
551 | goto out; | |
552 | #endif | |
1da177e4 LT |
553 | |
554 | while (len > 0) { | |
1da177e4 | 555 | int offset, bytes_to_copy; |
1da177e4 | 556 | |
9aea5a65 RM |
557 | if (fatal_signal_pending(current)) { |
558 | ret = -ERESTARTNOHAND; | |
559 | goto out; | |
560 | } | |
7993bc1f RM |
561 | cond_resched(); |
562 | ||
1da177e4 | 563 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
564 | if (offset == 0) |
565 | offset = PAGE_SIZE; | |
566 | ||
567 | bytes_to_copy = offset; | |
568 | if (bytes_to_copy > len) | |
569 | bytes_to_copy = len; | |
570 | ||
571 | offset -= bytes_to_copy; | |
572 | pos -= bytes_to_copy; | |
573 | str -= bytes_to_copy; | |
574 | len -= bytes_to_copy; | |
575 | ||
576 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
577 | struct page *page; | |
578 | ||
579 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 580 | if (!page) { |
b6a2fea3 | 581 | ret = -E2BIG; |
1da177e4 LT |
582 | goto out; |
583 | } | |
1da177e4 | 584 | |
b6a2fea3 | 585 | if (kmapped_page) { |
f358afc5 | 586 | flush_dcache_page(kmapped_page); |
1da177e4 | 587 | kunmap(kmapped_page); |
b6a2fea3 OW |
588 | put_arg_page(kmapped_page); |
589 | } | |
1da177e4 LT |
590 | kmapped_page = page; |
591 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
592 | kpos = pos & PAGE_MASK; |
593 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 594 | } |
b6a2fea3 | 595 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
596 | ret = -EFAULT; |
597 | goto out; | |
598 | } | |
1da177e4 LT |
599 | } |
600 | } | |
601 | ret = 0; | |
602 | out: | |
b6a2fea3 | 603 | if (kmapped_page) { |
f358afc5 | 604 | flush_dcache_page(kmapped_page); |
1da177e4 | 605 | kunmap(kmapped_page); |
b6a2fea3 OW |
606 | put_arg_page(kmapped_page); |
607 | } | |
1da177e4 LT |
608 | return ret; |
609 | } | |
610 | ||
611 | /* | |
986db2d1 | 612 | * Copy and argument/environment string from the kernel to the processes stack. |
1da177e4 | 613 | */ |
986db2d1 | 614 | int copy_string_kernel(const char *arg, struct linux_binprm *bprm) |
1da177e4 | 615 | { |
762a3af6 CH |
616 | int len = strnlen(arg, MAX_ARG_STRLEN) + 1 /* terminating NUL */; |
617 | unsigned long pos = bprm->p; | |
618 | ||
619 | if (len == 0) | |
620 | return -EFAULT; | |
621 | if (!valid_arg_len(bprm, len)) | |
622 | return -E2BIG; | |
623 | ||
624 | /* We're going to work our way backwards. */ | |
625 | arg += len; | |
626 | bprm->p -= len; | |
627 | if (IS_ENABLED(CONFIG_MMU) && bprm->p < bprm->argmin) | |
628 | return -E2BIG; | |
ba2d0162 | 629 | |
762a3af6 CH |
630 | while (len > 0) { |
631 | unsigned int bytes_to_copy = min_t(unsigned int, len, | |
632 | min_not_zero(offset_in_page(pos), PAGE_SIZE)); | |
633 | struct page *page; | |
634 | char *kaddr; | |
ba2d0162 | 635 | |
762a3af6 CH |
636 | pos -= bytes_to_copy; |
637 | arg -= bytes_to_copy; | |
638 | len -= bytes_to_copy; | |
ba2d0162 | 639 | |
762a3af6 CH |
640 | page = get_arg_page(bprm, pos, 1); |
641 | if (!page) | |
642 | return -E2BIG; | |
643 | kaddr = kmap_atomic(page); | |
644 | flush_arg_page(bprm, pos & PAGE_MASK, page); | |
645 | memcpy(kaddr + offset_in_page(pos), arg, bytes_to_copy); | |
f358afc5 | 646 | flush_dcache_page(page); |
762a3af6 CH |
647 | kunmap_atomic(kaddr); |
648 | put_arg_page(page); | |
649 | } | |
650 | ||
651 | return 0; | |
1da177e4 | 652 | } |
986db2d1 | 653 | EXPORT_SYMBOL(copy_string_kernel); |
1da177e4 | 654 | |
be619f7f EB |
655 | static int copy_strings_kernel(int argc, const char *const *argv, |
656 | struct linux_binprm *bprm) | |
657 | { | |
658 | while (argc-- > 0) { | |
659 | int ret = copy_string_kernel(argv[argc], bprm); | |
660 | if (ret < 0) | |
661 | return ret; | |
662 | if (fatal_signal_pending(current)) | |
663 | return -ERESTARTNOHAND; | |
664 | cond_resched(); | |
665 | } | |
666 | return 0; | |
667 | } | |
668 | ||
1da177e4 | 669 | #ifdef CONFIG_MMU |
b6a2fea3 | 670 | |
1da177e4 | 671 | /* |
b6a2fea3 OW |
672 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
673 | * the binfmt code determines where the new stack should reside, we shift it to | |
674 | * its final location. The process proceeds as follows: | |
1da177e4 | 675 | * |
b6a2fea3 OW |
676 | * 1) Use shift to calculate the new vma endpoints. |
677 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
678 | * arguments passed to subsequent functions are consistent. | |
679 | * 3) Move vma's page tables to the new range. | |
680 | * 4) Free up any cleared pgd range. | |
681 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 682 | */ |
b6a2fea3 | 683 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
684 | { |
685 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
686 | unsigned long old_start = vma->vm_start; |
687 | unsigned long old_end = vma->vm_end; | |
688 | unsigned long length = old_end - old_start; | |
689 | unsigned long new_start = old_start - shift; | |
690 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 691 | struct mmu_gather tlb; |
1da177e4 | 692 | |
b6a2fea3 | 693 | BUG_ON(new_start > new_end); |
1da177e4 | 694 | |
b6a2fea3 OW |
695 | /* |
696 | * ensure there are no vmas between where we want to go | |
697 | * and where we are | |
698 | */ | |
699 | if (vma != find_vma(mm, new_start)) | |
700 | return -EFAULT; | |
701 | ||
702 | /* | |
703 | * cover the whole range: [new_start, old_end) | |
704 | */ | |
5beb4930 RR |
705 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
706 | return -ENOMEM; | |
b6a2fea3 OW |
707 | |
708 | /* | |
709 | * move the page tables downwards, on failure we rely on | |
710 | * process cleanup to remove whatever mess we made. | |
711 | */ | |
712 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 713 | vma, new_start, length, false)) |
b6a2fea3 OW |
714 | return -ENOMEM; |
715 | ||
716 | lru_add_drain(); | |
a72afd87 | 717 | tlb_gather_mmu(&tlb, mm); |
b6a2fea3 OW |
718 | if (new_end > old_start) { |
719 | /* | |
720 | * when the old and new regions overlap clear from new_end. | |
721 | */ | |
d16dfc55 | 722 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 723 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
724 | } else { |
725 | /* | |
726 | * otherwise, clean from old_start; this is done to not touch | |
727 | * the address space in [new_end, old_start) some architectures | |
728 | * have constraints on va-space that make this illegal (IA64) - | |
729 | * for the others its just a little faster. | |
730 | */ | |
d16dfc55 | 731 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 732 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 733 | } |
ae8eba8b | 734 | tlb_finish_mmu(&tlb); |
b6a2fea3 OW |
735 | |
736 | /* | |
5beb4930 | 737 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
738 | */ |
739 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
740 | ||
741 | return 0; | |
1da177e4 LT |
742 | } |
743 | ||
b6a2fea3 OW |
744 | /* |
745 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
746 | * the stack is optionally relocated, and some extra space is added. | |
747 | */ | |
1da177e4 LT |
748 | int setup_arg_pages(struct linux_binprm *bprm, |
749 | unsigned long stack_top, | |
750 | int executable_stack) | |
751 | { | |
b6a2fea3 OW |
752 | unsigned long ret; |
753 | unsigned long stack_shift; | |
1da177e4 | 754 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
755 | struct vm_area_struct *vma = bprm->vma; |
756 | struct vm_area_struct *prev = NULL; | |
757 | unsigned long vm_flags; | |
758 | unsigned long stack_base; | |
803bf5ec MN |
759 | unsigned long stack_size; |
760 | unsigned long stack_expand; | |
761 | unsigned long rlim_stack; | |
1da177e4 LT |
762 | |
763 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 764 | /* Limit stack size */ |
c31dbb14 | 765 | stack_base = bprm->rlim_stack.rlim_max; |
22ee3ea5 HD |
766 | |
767 | stack_base = calc_max_stack_size(stack_base); | |
1da177e4 | 768 | |
d045c77c HD |
769 | /* Add space for stack randomization. */ |
770 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
771 | ||
b6a2fea3 OW |
772 | /* Make sure we didn't let the argument array grow too large. */ |
773 | if (vma->vm_end - vma->vm_start > stack_base) | |
774 | return -ENOMEM; | |
1da177e4 | 775 | |
b6a2fea3 | 776 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 777 | |
b6a2fea3 OW |
778 | stack_shift = vma->vm_start - stack_base; |
779 | mm->arg_start = bprm->p - stack_shift; | |
780 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 781 | #else |
b6a2fea3 OW |
782 | stack_top = arch_align_stack(stack_top); |
783 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
784 | |
785 | if (unlikely(stack_top < mmap_min_addr) || | |
786 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
787 | return -ENOMEM; | |
788 | ||
b6a2fea3 OW |
789 | stack_shift = vma->vm_end - stack_top; |
790 | ||
791 | bprm->p -= stack_shift; | |
1da177e4 | 792 | mm->arg_start = bprm->p; |
1da177e4 LT |
793 | #endif |
794 | ||
1da177e4 | 795 | if (bprm->loader) |
b6a2fea3 OW |
796 | bprm->loader -= stack_shift; |
797 | bprm->exec -= stack_shift; | |
1da177e4 | 798 | |
d8ed45c5 | 799 | if (mmap_write_lock_killable(mm)) |
f268dfe9 MH |
800 | return -EINTR; |
801 | ||
96a8e13e | 802 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
803 | |
804 | /* | |
805 | * Adjust stack execute permissions; explicitly enable for | |
806 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
807 | * (arch default) otherwise. | |
808 | */ | |
809 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
810 | vm_flags |= VM_EXEC; | |
811 | else if (executable_stack == EXSTACK_DISABLE_X) | |
812 | vm_flags &= ~VM_EXEC; | |
813 | vm_flags |= mm->def_flags; | |
a8bef8ff | 814 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
815 | |
816 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
817 | vm_flags); | |
818 | if (ret) | |
819 | goto out_unlock; | |
820 | BUG_ON(prev != vma); | |
821 | ||
47a2ebb7 AD |
822 | if (unlikely(vm_flags & VM_EXEC)) { |
823 | pr_warn_once("process '%pD4' started with executable stack\n", | |
824 | bprm->file); | |
825 | } | |
826 | ||
b6a2fea3 OW |
827 | /* Move stack pages down in memory. */ |
828 | if (stack_shift) { | |
829 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
830 | if (ret) |
831 | goto out_unlock; | |
1da177e4 LT |
832 | } |
833 | ||
a8bef8ff MG |
834 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
835 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
836 | ||
5ef097dd | 837 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
838 | stack_size = vma->vm_end - vma->vm_start; |
839 | /* | |
840 | * Align this down to a page boundary as expand_stack | |
841 | * will align it up. | |
842 | */ | |
c31dbb14 | 843 | rlim_stack = bprm->rlim_stack.rlim_cur & PAGE_MASK; |
b6a2fea3 | 844 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
845 | if (stack_size + stack_expand > rlim_stack) |
846 | stack_base = vma->vm_start + rlim_stack; | |
847 | else | |
848 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 849 | #else |
803bf5ec MN |
850 | if (stack_size + stack_expand > rlim_stack) |
851 | stack_base = vma->vm_end - rlim_stack; | |
852 | else | |
853 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 854 | #endif |
3af9e859 | 855 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
856 | ret = expand_stack(vma, stack_base); |
857 | if (ret) | |
858 | ret = -EFAULT; | |
859 | ||
860 | out_unlock: | |
d8ed45c5 | 861 | mmap_write_unlock(mm); |
fc63cf23 | 862 | return ret; |
1da177e4 | 863 | } |
1da177e4 LT |
864 | EXPORT_SYMBOL(setup_arg_pages); |
865 | ||
7e7ec6a9 NP |
866 | #else |
867 | ||
868 | /* | |
869 | * Transfer the program arguments and environment from the holding pages | |
870 | * onto the stack. The provided stack pointer is adjusted accordingly. | |
871 | */ | |
872 | int transfer_args_to_stack(struct linux_binprm *bprm, | |
873 | unsigned long *sp_location) | |
874 | { | |
875 | unsigned long index, stop, sp; | |
876 | int ret = 0; | |
877 | ||
878 | stop = bprm->p >> PAGE_SHIFT; | |
879 | sp = *sp_location; | |
880 | ||
881 | for (index = MAX_ARG_PAGES - 1; index >= stop; index--) { | |
882 | unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0; | |
883 | char *src = kmap(bprm->page[index]) + offset; | |
884 | sp -= PAGE_SIZE - offset; | |
885 | if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0) | |
886 | ret = -EFAULT; | |
887 | kunmap(bprm->page[index]); | |
888 | if (ret) | |
889 | goto out; | |
890 | } | |
891 | ||
892 | *sp_location = sp; | |
893 | ||
894 | out: | |
895 | return ret; | |
896 | } | |
897 | EXPORT_SYMBOL(transfer_args_to_stack); | |
898 | ||
1da177e4 LT |
899 | #endif /* CONFIG_MMU */ |
900 | ||
51f39a1f | 901 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 902 | { |
1da177e4 | 903 | struct file *file; |
e56b6a5d | 904 | int err; |
51f39a1f | 905 | struct open_flags open_exec_flags = { |
47c805dc | 906 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
62fb4a15 | 907 | .acc_mode = MAY_EXEC, |
f9652e10 AV |
908 | .intent = LOOKUP_OPEN, |
909 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 910 | }; |
1da177e4 | 911 | |
51f39a1f DD |
912 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
913 | return ERR_PTR(-EINVAL); | |
914 | if (flags & AT_SYMLINK_NOFOLLOW) | |
915 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
916 | if (flags & AT_EMPTY_PATH) | |
917 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
918 | ||
919 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 920 | if (IS_ERR(file)) |
e56b6a5d CH |
921 | goto out; |
922 | ||
633fb6ac KC |
923 | /* |
924 | * may_open() has already checked for this, so it should be | |
925 | * impossible to trip now. But we need to be extra cautious | |
926 | * and check again at the very end too. | |
927 | */ | |
e56b6a5d | 928 | err = -EACCES; |
0fd338b2 KC |
929 | if (WARN_ON_ONCE(!S_ISREG(file_inode(file)->i_mode) || |
930 | path_noexec(&file->f_path))) | |
6e8341a1 | 931 | goto exit; |
e56b6a5d CH |
932 | |
933 | err = deny_write_access(file); | |
6e8341a1 AV |
934 | if (err) |
935 | goto exit; | |
1da177e4 | 936 | |
51f39a1f DD |
937 | if (name->name[0] != '\0') |
938 | fsnotify_open(file); | |
939 | ||
6e8341a1 | 940 | out: |
e56b6a5d CH |
941 | return file; |
942 | ||
6e8341a1 AV |
943 | exit: |
944 | fput(file); | |
e56b6a5d CH |
945 | return ERR_PTR(err); |
946 | } | |
c4ad8f98 LT |
947 | |
948 | struct file *open_exec(const char *name) | |
949 | { | |
51689104 PM |
950 | struct filename *filename = getname_kernel(name); |
951 | struct file *f = ERR_CAST(filename); | |
952 | ||
953 | if (!IS_ERR(filename)) { | |
954 | f = do_open_execat(AT_FDCWD, filename, 0); | |
955 | putname(filename); | |
956 | } | |
957 | return f; | |
c4ad8f98 | 958 | } |
1da177e4 LT |
959 | EXPORT_SYMBOL(open_exec); |
960 | ||
48304f79 CH |
961 | #if defined(CONFIG_HAVE_AOUT) || defined(CONFIG_BINFMT_FLAT) || \ |
962 | defined(CONFIG_BINFMT_ELF_FDPIC) | |
3dc20cb2 AV |
963 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
964 | { | |
ec695579 | 965 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 | 966 | if (res > 0) |
bce2b68b | 967 | flush_icache_user_range(addr, addr + len); |
3dc20cb2 AV |
968 | return res; |
969 | } | |
970 | EXPORT_SYMBOL(read_code); | |
48304f79 | 971 | #endif |
3dc20cb2 | 972 | |
eea96732 EB |
973 | /* |
974 | * Maps the mm_struct mm into the current task struct. | |
f7cfd871 EB |
975 | * On success, this function returns with exec_update_lock |
976 | * held for writing. | |
eea96732 | 977 | */ |
1da177e4 LT |
978 | static int exec_mmap(struct mm_struct *mm) |
979 | { | |
980 | struct task_struct *tsk; | |
615d6e87 | 981 | struct mm_struct *old_mm, *active_mm; |
eea96732 | 982 | int ret; |
1da177e4 LT |
983 | |
984 | /* Notify parent that we're no longer interested in the old VM */ | |
985 | tsk = current; | |
986 | old_mm = current->mm; | |
4610ba7a | 987 | exec_mm_release(tsk, old_mm); |
a28bf136 EB |
988 | if (old_mm) |
989 | sync_mm_rss(old_mm); | |
1da177e4 | 990 | |
f7cfd871 | 991 | ret = down_write_killable(&tsk->signal->exec_update_lock); |
eea96732 EB |
992 | if (ret) |
993 | return ret; | |
994 | ||
1da177e4 LT |
995 | if (old_mm) { |
996 | /* | |
997 | * Make sure that if there is a core dump in progress | |
998 | * for the old mm, we get out and die instead of going | |
c1e8d7c6 | 999 | * through with the exec. We must hold mmap_lock around |
999d9fc1 | 1000 | * checking core_state and changing tsk->mm. |
1da177e4 | 1001 | */ |
d8ed45c5 | 1002 | mmap_read_lock(old_mm); |
999d9fc1 | 1003 | if (unlikely(old_mm->core_state)) { |
d8ed45c5 | 1004 | mmap_read_unlock(old_mm); |
f7cfd871 | 1005 | up_write(&tsk->signal->exec_update_lock); |
1da177e4 LT |
1006 | return -EINTR; |
1007 | } | |
1008 | } | |
eea96732 | 1009 | |
1da177e4 | 1010 | task_lock(tsk); |
227a4aad | 1011 | membarrier_exec_mmap(mm); |
d53c3dfb NP |
1012 | |
1013 | local_irq_disable(); | |
1014 | active_mm = tsk->active_mm; | |
1da177e4 | 1015 | tsk->active_mm = mm; |
d53c3dfb NP |
1016 | tsk->mm = mm; |
1017 | /* | |
1018 | * This prevents preemption while active_mm is being loaded and | |
1019 | * it and mm are being updated, which could cause problems for | |
1020 | * lazy tlb mm refcounting when these are updated by context | |
1021 | * switches. Not all architectures can handle irqs off over | |
1022 | * activate_mm yet. | |
1023 | */ | |
1024 | if (!IS_ENABLED(CONFIG_ARCH_WANT_IRQS_OFF_ACTIVATE_MM)) | |
1025 | local_irq_enable(); | |
1da177e4 | 1026 | activate_mm(active_mm, mm); |
d53c3dfb NP |
1027 | if (IS_ENABLED(CONFIG_ARCH_WANT_IRQS_OFF_ACTIVATE_MM)) |
1028 | local_irq_enable(); | |
615d6e87 DB |
1029 | tsk->mm->vmacache_seqnum = 0; |
1030 | vmacache_flush(tsk); | |
1da177e4 | 1031 | task_unlock(tsk); |
1da177e4 | 1032 | if (old_mm) { |
d8ed45c5 | 1033 | mmap_read_unlock(old_mm); |
7dddb12c | 1034 | BUG_ON(active_mm != old_mm); |
701085b2 | 1035 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 1036 | mm_update_next_owner(old_mm); |
1da177e4 LT |
1037 | mmput(old_mm); |
1038 | return 0; | |
1039 | } | |
1040 | mmdrop(active_mm); | |
1041 | return 0; | |
1042 | } | |
1043 | ||
858119e1 | 1044 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
1045 | { |
1046 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 1047 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 1048 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 1049 | |
aafe6c2a | 1050 | if (thread_group_empty(tsk)) |
1da177e4 LT |
1051 | goto no_thread_group; |
1052 | ||
1053 | /* | |
1054 | * Kill all other threads in the thread group. | |
1da177e4 | 1055 | */ |
1da177e4 | 1056 | spin_lock_irq(lock); |
ed5d2cac | 1057 | if (signal_group_exit(sig)) { |
1da177e4 LT |
1058 | /* |
1059 | * Another group action in progress, just | |
1060 | * return so that the signal is processed. | |
1061 | */ | |
1062 | spin_unlock_irq(lock); | |
1da177e4 LT |
1063 | return -EAGAIN; |
1064 | } | |
d344193a | 1065 | |
ed5d2cac | 1066 | sig->group_exit_task = tsk; |
d344193a ON |
1067 | sig->notify_count = zap_other_threads(tsk); |
1068 | if (!thread_group_leader(tsk)) | |
1069 | sig->notify_count--; | |
1da177e4 | 1070 | |
d344193a | 1071 | while (sig->notify_count) { |
d5bbd43d | 1072 | __set_current_state(TASK_KILLABLE); |
1da177e4 | 1073 | spin_unlock_irq(lock); |
a72173ec | 1074 | schedule(); |
08d405c8 | 1075 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1076 | goto killed; |
1da177e4 LT |
1077 | spin_lock_irq(lock); |
1078 | } | |
1da177e4 LT |
1079 | spin_unlock_irq(lock); |
1080 | ||
1081 | /* | |
1082 | * At this point all other threads have exited, all we have to | |
1083 | * do is to wait for the thread group leader to become inactive, | |
1084 | * and to assume its PID: | |
1085 | */ | |
aafe6c2a | 1086 | if (!thread_group_leader(tsk)) { |
8187926b | 1087 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 1088 | |
6db840fa | 1089 | for (;;) { |
780de9dd | 1090 | cgroup_threadgroup_change_begin(tsk); |
6db840fa | 1091 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
1092 | /* |
1093 | * Do this under tasklist_lock to ensure that | |
1094 | * exit_notify() can't miss ->group_exit_task | |
1095 | */ | |
1096 | sig->notify_count = -1; | |
6db840fa ON |
1097 | if (likely(leader->exit_state)) |
1098 | break; | |
d5bbd43d | 1099 | __set_current_state(TASK_KILLABLE); |
6db840fa | 1100 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1101 | cgroup_threadgroup_change_end(tsk); |
a72173ec | 1102 | schedule(); |
08d405c8 | 1103 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1104 | goto killed; |
6db840fa | 1105 | } |
1da177e4 | 1106 | |
f5e90281 RM |
1107 | /* |
1108 | * The only record we have of the real-time age of a | |
1109 | * process, regardless of execs it's done, is start_time. | |
1110 | * All the past CPU time is accumulated in signal_struct | |
1111 | * from sister threads now dead. But in this non-leader | |
1112 | * exec, nothing survives from the original leader thread, | |
1113 | * whose birth marks the true age of this process now. | |
1114 | * When we take on its identity by switching to its PID, we | |
1115 | * also take its birthdate (always earlier than our own). | |
1116 | */ | |
aafe6c2a | 1117 | tsk->start_time = leader->start_time; |
cf25e24d | 1118 | tsk->start_boottime = leader->start_boottime; |
f5e90281 | 1119 | |
bac0abd6 | 1120 | BUG_ON(!same_thread_group(leader, tsk)); |
1da177e4 LT |
1121 | /* |
1122 | * An exec() starts a new thread group with the | |
1123 | * TGID of the previous thread group. Rehash the | |
1124 | * two threads with a switched PID, and release | |
1125 | * the former thread group leader: | |
1126 | */ | |
d73d6529 EB |
1127 | |
1128 | /* Become a process group leader with the old leader's pid. | |
c18258c6 | 1129 | * The old leader becomes a thread of the this thread group. |
d73d6529 | 1130 | */ |
6b03d130 | 1131 | exchange_tids(tsk, leader); |
6883f81a | 1132 | transfer_pid(leader, tsk, PIDTYPE_TGID); |
aafe6c2a EB |
1133 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
1134 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 1135 | |
aafe6c2a | 1136 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1137 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1138 | |
aafe6c2a EB |
1139 | tsk->group_leader = tsk; |
1140 | leader->group_leader = tsk; | |
de12a787 | 1141 | |
aafe6c2a | 1142 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1143 | leader->exit_signal = -1; |
962b564c ON |
1144 | |
1145 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1146 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1147 | |
1148 | /* | |
1149 | * We are going to release_task()->ptrace_unlink() silently, | |
1150 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1151 | * the tracer wont't block again waiting for this thread. | |
1152 | */ | |
1153 | if (unlikely(leader->ptrace)) | |
1154 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1155 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1156 | cgroup_threadgroup_change_end(tsk); |
8187926b ON |
1157 | |
1158 | release_task(leader); | |
ed5d2cac | 1159 | } |
1da177e4 | 1160 | |
6db840fa ON |
1161 | sig->group_exit_task = NULL; |
1162 | sig->notify_count = 0; | |
1da177e4 LT |
1163 | |
1164 | no_thread_group: | |
e6368253 ON |
1165 | /* we have changed execution domain */ |
1166 | tsk->exit_signal = SIGCHLD; | |
1167 | ||
02169155 EB |
1168 | BUG_ON(!thread_group_leader(tsk)); |
1169 | return 0; | |
1170 | ||
1171 | killed: | |
1172 | /* protects against exit_notify() and __exit_signal() */ | |
1173 | read_lock(&tasklist_lock); | |
1174 | sig->group_exit_task = NULL; | |
1175 | sig->notify_count = 0; | |
1176 | read_unlock(&tasklist_lock); | |
1177 | return -EAGAIN; | |
1178 | } | |
1179 | ||
1180 | ||
7a60ef48 EB |
1181 | /* |
1182 | * This function makes sure the current process has its own signal table, | |
1183 | * so that flush_signal_handlers can later reset the handlers without | |
1184 | * disturbing other processes. (Other processes might share the signal | |
1185 | * table via the CLONE_SIGHAND option to clone().) | |
1186 | */ | |
02169155 EB |
1187 | static int unshare_sighand(struct task_struct *me) |
1188 | { | |
1189 | struct sighand_struct *oldsighand = me->sighand; | |
329f7dba | 1190 | |
d036bda7 | 1191 | if (refcount_read(&oldsighand->count) != 1) { |
b2c903b8 | 1192 | struct sighand_struct *newsighand; |
1da177e4 | 1193 | /* |
b2c903b8 ON |
1194 | * This ->sighand is shared with the CLONE_SIGHAND |
1195 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1196 | */ |
b2c903b8 ON |
1197 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1198 | if (!newsighand) | |
1199 | return -ENOMEM; | |
1200 | ||
d036bda7 | 1201 | refcount_set(&newsighand->count, 1); |
1da177e4 LT |
1202 | memcpy(newsighand->action, oldsighand->action, |
1203 | sizeof(newsighand->action)); | |
1204 | ||
1205 | write_lock_irq(&tasklist_lock); | |
1206 | spin_lock(&oldsighand->siglock); | |
02169155 | 1207 | rcu_assign_pointer(me->sighand, newsighand); |
1da177e4 LT |
1208 | spin_unlock(&oldsighand->siglock); |
1209 | write_unlock_irq(&tasklist_lock); | |
1210 | ||
fba2afaa | 1211 | __cleanup_sighand(oldsighand); |
1da177e4 | 1212 | } |
1da177e4 LT |
1213 | return 0; |
1214 | } | |
0840a90d | 1215 | |
3756f640 | 1216 | char *__get_task_comm(char *buf, size_t buf_size, struct task_struct *tsk) |
1da177e4 | 1217 | { |
1da177e4 | 1218 | task_lock(tsk); |
3756f640 | 1219 | strncpy(buf, tsk->comm, buf_size); |
1da177e4 | 1220 | task_unlock(tsk); |
59714d65 | 1221 | return buf; |
1da177e4 | 1222 | } |
3756f640 | 1223 | EXPORT_SYMBOL_GPL(__get_task_comm); |
1da177e4 | 1224 | |
6a6d27de AV |
1225 | /* |
1226 | * These functions flushes out all traces of the currently running executable | |
1227 | * so that a new one can be started | |
1228 | */ | |
1229 | ||
82b89778 | 1230 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1231 | { |
1232 | task_lock(tsk); | |
43d2b113 | 1233 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1234 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1235 | task_unlock(tsk); | |
82b89778 | 1236 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1237 | } |
1238 | ||
a9208e42 KC |
1239 | /* |
1240 | * Calling this is the point of no return. None of the failures will be | |
1241 | * seen by userspace since either the process is already taking a fatal | |
1242 | * signal (via de_thread() or coredump), or will have SEGV raised | |
13c432b5 | 1243 | * (after exec_mmap()) by search_binary_handler (see below). |
a9208e42 | 1244 | */ |
2388777a | 1245 | int begin_new_exec(struct linux_binprm * bprm) |
1da177e4 | 1246 | { |
2ca7be7d | 1247 | struct task_struct *me = current; |
221af7f8 | 1248 | int retval; |
1da177e4 | 1249 | |
56305aa9 EB |
1250 | /* Once we are committed compute the creds */ |
1251 | retval = bprm_creds_from_file(bprm); | |
1252 | if (retval) | |
1253 | return retval; | |
1254 | ||
6834e0bb EB |
1255 | /* |
1256 | * Ensure all future errors are fatal. | |
1257 | */ | |
1258 | bprm->point_of_no_return = true; | |
1259 | ||
1da177e4 | 1260 | /* |
02169155 | 1261 | * Make this the only thread in the thread group. |
1da177e4 | 1262 | */ |
2ca7be7d | 1263 | retval = de_thread(me); |
1da177e4 LT |
1264 | if (retval) |
1265 | goto out; | |
1266 | ||
9ee1206d EB |
1267 | /* |
1268 | * Cancel any io_uring activity across execve | |
1269 | */ | |
1270 | io_uring_task_cancel(); | |
1271 | ||
b6043501 | 1272 | /* Ensure the files table is not shared. */ |
1f702603 | 1273 | retval = unshare_files(); |
b6043501 EB |
1274 | if (retval) |
1275 | goto out; | |
b6043501 | 1276 | |
6e399cd1 DB |
1277 | /* |
1278 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1279 | * not visibile until then. This also enables the update | |
1280 | * to be lockless. | |
1281 | */ | |
fe69d560 DH |
1282 | retval = set_mm_exe_file(bprm->mm, bprm->file); |
1283 | if (retval) | |
1284 | goto out; | |
6e399cd1 | 1285 | |
b8a61c9e | 1286 | /* If the binary is not readable then enforce mm->dumpable=0 */ |
f87d1c95 | 1287 | would_dump(bprm, bprm->file); |
b8a61c9e EB |
1288 | if (bprm->have_execfd) |
1289 | would_dump(bprm, bprm->executable); | |
f87d1c95 | 1290 | |
1da177e4 LT |
1291 | /* |
1292 | * Release all of the old mmap stuff | |
1293 | */ | |
3c77f845 | 1294 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1295 | retval = exec_mmap(bprm->mm); |
1296 | if (retval) | |
fd8328be | 1297 | goto out; |
1da177e4 | 1298 | |
a9208e42 | 1299 | bprm->mm = NULL; |
7ab02af4 | 1300 | |
ccf0fa6b | 1301 | #ifdef CONFIG_POSIX_TIMERS |
3ed572c8 TLSC |
1302 | spin_lock_irq(&me->sighand->siglock); |
1303 | posix_cpu_timers_exit(me); | |
1304 | spin_unlock_irq(&me->sighand->siglock); | |
ccf0fa6b EB |
1305 | exit_itimers(me->signal); |
1306 | flush_itimer_signals(); | |
1307 | #endif | |
1308 | ||
1309 | /* | |
1310 | * Make the signal table private. | |
1311 | */ | |
1312 | retval = unshare_sighand(me); | |
1313 | if (retval) | |
89826cce | 1314 | goto out_unlock; |
ccf0fa6b | 1315 | |
fe814175 CH |
1316 | /* |
1317 | * Ensure that the uaccess routines can actually operate on userspace | |
1318 | * pointers: | |
1319 | */ | |
1320 | force_uaccess_begin(); | |
1321 | ||
2ca7be7d | 1322 | me->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
b88fae64 | 1323 | PF_NOFREEZE | PF_NO_SETAFFINITY); |
7ab02af4 | 1324 | flush_thread(); |
2ca7be7d | 1325 | me->personality &= ~bprm->per_clear; |
7ab02af4 | 1326 | |
1446e1df GKB |
1327 | clear_syscall_work_syscall_user_dispatch(me); |
1328 | ||
613cc2b6 AS |
1329 | /* |
1330 | * We have to apply CLOEXEC before we change whether the process is | |
1331 | * dumpable (in setup_new_exec) to avoid a race with a process in userspace | |
1332 | * trying to access the should-be-closed file descriptors of a process | |
1333 | * undergoing exec(2). | |
1334 | */ | |
2ca7be7d | 1335 | do_close_on_exec(me->files); |
f84df2a6 | 1336 | |
64701dee | 1337 | if (bprm->secureexec) { |
fe8993b3 | 1338 | /* Make sure parent cannot signal privileged process. */ |
7d503feb | 1339 | me->pdeath_signal = 0; |
fe8993b3 | 1340 | |
64701dee KC |
1341 | /* |
1342 | * For secureexec, reset the stack limit to sane default to | |
1343 | * avoid bad behavior from the prior rlimits. This has to | |
1344 | * happen before arch_pick_mmap_layout(), which examines | |
1345 | * RLIMIT_STACK, but after the point of no return to avoid | |
779f4e1c | 1346 | * needing to clean up the change on failure. |
64701dee | 1347 | */ |
c31dbb14 KC |
1348 | if (bprm->rlim_stack.rlim_cur > _STK_LIM) |
1349 | bprm->rlim_stack.rlim_cur = _STK_LIM; | |
64701dee KC |
1350 | } |
1351 | ||
7d503feb | 1352 | me->sas_ss_sp = me->sas_ss_size = 0; |
1da177e4 | 1353 | |
e816c201 KC |
1354 | /* |
1355 | * Figure out dumpability. Note that this checking only of current | |
1356 | * is wrong, but userspace depends on it. This should be testing | |
1357 | * bprm->secureexec instead. | |
1358 | */ | |
473d8963 | 1359 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP || |
e816c201 KC |
1360 | !(uid_eq(current_euid(), current_uid()) && |
1361 | gid_eq(current_egid(), current_gid()))) | |
6c5d5238 | 1362 | set_dumpable(current->mm, suid_dumpable); |
473d8963 KC |
1363 | else |
1364 | set_dumpable(current->mm, SUID_DUMP_USER); | |
d6e71144 | 1365 | |
e041e328 | 1366 | perf_event_exec(); |
7d503feb | 1367 | __set_task_comm(me, kbasename(bprm->filename), true); |
1da177e4 | 1368 | |
1da177e4 LT |
1369 | /* An exec changes our domain. We are no longer part of the thread |
1370 | group */ | |
7d503feb EB |
1371 | WRITE_ONCE(me->self_exec_id, me->self_exec_id + 1); |
1372 | flush_signal_handlers(me, 0); | |
96ecee29 | 1373 | |
905ae01c AG |
1374 | retval = set_cred_ucounts(bprm->cred); |
1375 | if (retval < 0) | |
1376 | goto out_unlock; | |
1377 | ||
96ecee29 EB |
1378 | /* |
1379 | * install the new credentials for this executable | |
1380 | */ | |
1381 | security_bprm_committing_creds(bprm); | |
1382 | ||
1383 | commit_creds(bprm->cred); | |
1384 | bprm->cred = NULL; | |
1385 | ||
1386 | /* | |
1387 | * Disable monitoring for regular users | |
1388 | * when executing setuid binaries. Must | |
1389 | * wait until new credentials are committed | |
1390 | * by commit_creds() above | |
1391 | */ | |
7d503feb EB |
1392 | if (get_dumpable(me->mm) != SUID_DUMP_USER) |
1393 | perf_event_exit_task(me); | |
96ecee29 EB |
1394 | /* |
1395 | * cred_guard_mutex must be held at least to this point to prevent | |
1396 | * ptrace_attach() from altering our determination of the task's | |
1397 | * credentials; any time after this it may be unlocked. | |
1398 | */ | |
1399 | security_bprm_committed_creds(bprm); | |
b8a61c9e EB |
1400 | |
1401 | /* Pass the opened binary to the interpreter. */ | |
1402 | if (bprm->have_execfd) { | |
1403 | retval = get_unused_fd_flags(0); | |
1404 | if (retval < 0) | |
1405 | goto out_unlock; | |
1406 | fd_install(retval, bprm->executable); | |
1407 | bprm->executable = NULL; | |
1408 | bprm->execfd = retval; | |
1409 | } | |
221af7f8 LT |
1410 | return 0; |
1411 | ||
df9e4d2c | 1412 | out_unlock: |
f7cfd871 | 1413 | up_write(&me->signal->exec_update_lock); |
221af7f8 LT |
1414 | out: |
1415 | return retval; | |
1416 | } | |
2388777a | 1417 | EXPORT_SYMBOL(begin_new_exec); |
221af7f8 | 1418 | |
1b5d783c AV |
1419 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1420 | { | |
f84df2a6 | 1421 | struct inode *inode = file_inode(file); |
435ac621 CB |
1422 | struct user_namespace *mnt_userns = file_mnt_user_ns(file); |
1423 | if (inode_permission(mnt_userns, inode, MAY_READ) < 0) { | |
f84df2a6 | 1424 | struct user_namespace *old, *user_ns; |
1b5d783c | 1425 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
f84df2a6 EB |
1426 | |
1427 | /* Ensure mm->user_ns contains the executable */ | |
1428 | user_ns = old = bprm->mm->user_ns; | |
1429 | while ((user_ns != &init_user_ns) && | |
435ac621 | 1430 | !privileged_wrt_inode_uidgid(user_ns, mnt_userns, inode)) |
f84df2a6 EB |
1431 | user_ns = user_ns->parent; |
1432 | ||
1433 | if (old != user_ns) { | |
1434 | bprm->mm->user_ns = get_user_ns(user_ns); | |
1435 | put_user_ns(old); | |
1436 | } | |
1437 | } | |
1b5d783c AV |
1438 | } |
1439 | EXPORT_SYMBOL(would_dump); | |
1440 | ||
221af7f8 LT |
1441 | void setup_new_exec(struct linux_binprm * bprm) |
1442 | { | |
df9e4d2c EB |
1443 | /* Setup things that can depend upon the personality */ |
1444 | struct task_struct *me = current; | |
1da177e4 | 1445 | |
df9e4d2c | 1446 | arch_pick_mmap_layout(me->mm, &bprm->rlim_stack); |
d6e71144 | 1447 | |
e9ea1e7f | 1448 | arch_setup_new_exec(); |
1da177e4 | 1449 | |
0551fbd2 BH |
1450 | /* Set the new mm task size. We have to do that late because it may |
1451 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1452 | * some architectures like powerpc | |
1453 | */ | |
df9e4d2c | 1454 | me->mm->task_size = TASK_SIZE; |
f7cfd871 | 1455 | up_write(&me->signal->exec_update_lock); |
7d503feb | 1456 | mutex_unlock(&me->signal->cred_guard_mutex); |
1da177e4 | 1457 | } |
221af7f8 | 1458 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1459 | |
b8383831 KC |
1460 | /* Runs immediately before start_thread() takes over. */ |
1461 | void finalize_exec(struct linux_binprm *bprm) | |
1462 | { | |
c31dbb14 KC |
1463 | /* Store any stack rlimit changes before starting thread. */ |
1464 | task_lock(current->group_leader); | |
1465 | current->signal->rlim[RLIMIT_STACK] = bprm->rlim_stack; | |
1466 | task_unlock(current->group_leader); | |
b8383831 KC |
1467 | } |
1468 | EXPORT_SYMBOL(finalize_exec); | |
1469 | ||
a2a8474c ON |
1470 | /* |
1471 | * Prepare credentials and lock ->cred_guard_mutex. | |
96ecee29 | 1472 | * setup_new_exec() commits the new creds and drops the lock. |
3d742d4b | 1473 | * Or, if exec fails before, free_bprm() should release ->cred |
a2a8474c ON |
1474 | * and unlock. |
1475 | */ | |
4addd264 | 1476 | static int prepare_bprm_creds(struct linux_binprm *bprm) |
a2a8474c | 1477 | { |
9b1bf12d | 1478 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1479 | return -ERESTARTNOINTR; |
1480 | ||
1481 | bprm->cred = prepare_exec_creds(); | |
1482 | if (likely(bprm->cred)) | |
1483 | return 0; | |
1484 | ||
9b1bf12d | 1485 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1486 | return -ENOMEM; |
1487 | } | |
1488 | ||
c4ad8f98 | 1489 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c | 1490 | { |
f18ac551 EB |
1491 | if (bprm->mm) { |
1492 | acct_arg_size(bprm, 0); | |
1493 | mmput(bprm->mm); | |
1494 | } | |
a2a8474c ON |
1495 | free_arg_pages(bprm); |
1496 | if (bprm->cred) { | |
9b1bf12d | 1497 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1498 | abort_creds(bprm->cred); |
1499 | } | |
63e46b95 ON |
1500 | if (bprm->file) { |
1501 | allow_write_access(bprm->file); | |
1502 | fput(bprm->file); | |
1503 | } | |
b8a61c9e EB |
1504 | if (bprm->executable) |
1505 | fput(bprm->executable); | |
b66c5984 KC |
1506 | /* If a binfmt changed the interp, free it. */ |
1507 | if (bprm->interp != bprm->filename) | |
1508 | kfree(bprm->interp); | |
60d9ad1d | 1509 | kfree(bprm->fdpath); |
a2a8474c ON |
1510 | kfree(bprm); |
1511 | } | |
1512 | ||
60d9ad1d | 1513 | static struct linux_binprm *alloc_bprm(int fd, struct filename *filename) |
0a8f36eb EB |
1514 | { |
1515 | struct linux_binprm *bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); | |
60d9ad1d | 1516 | int retval = -ENOMEM; |
0a8f36eb | 1517 | if (!bprm) |
60d9ad1d EB |
1518 | goto out; |
1519 | ||
1520 | if (fd == AT_FDCWD || filename->name[0] == '/') { | |
1521 | bprm->filename = filename->name; | |
1522 | } else { | |
1523 | if (filename->name[0] == '\0') | |
1524 | bprm->fdpath = kasprintf(GFP_KERNEL, "/dev/fd/%d", fd); | |
1525 | else | |
1526 | bprm->fdpath = kasprintf(GFP_KERNEL, "/dev/fd/%d/%s", | |
1527 | fd, filename->name); | |
1528 | if (!bprm->fdpath) | |
1529 | goto out_free; | |
1530 | ||
1531 | bprm->filename = bprm->fdpath; | |
1532 | } | |
1533 | bprm->interp = bprm->filename; | |
f18ac551 EB |
1534 | |
1535 | retval = bprm_mm_init(bprm); | |
1536 | if (retval) | |
1537 | goto out_free; | |
0a8f36eb | 1538 | return bprm; |
60d9ad1d EB |
1539 | |
1540 | out_free: | |
1541 | free_bprm(bprm); | |
1542 | out: | |
1543 | return ERR_PTR(retval); | |
0a8f36eb EB |
1544 | } |
1545 | ||
c2315c18 | 1546 | int bprm_change_interp(const char *interp, struct linux_binprm *bprm) |
b66c5984 KC |
1547 | { |
1548 | /* If a binfmt changed the interp, free it first. */ | |
1549 | if (bprm->interp != bprm->filename) | |
1550 | kfree(bprm->interp); | |
1551 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1552 | if (!bprm->interp) | |
1553 | return -ENOMEM; | |
1554 | return 0; | |
1555 | } | |
1556 | EXPORT_SYMBOL(bprm_change_interp); | |
1557 | ||
a6f76f23 DH |
1558 | /* |
1559 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1560 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1561 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1562 | */ |
9e00cdb0 | 1563 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1564 | { |
0bf2f3ae | 1565 | struct task_struct *p = current, *t; |
f1191b50 | 1566 | unsigned n_fs; |
a6f76f23 | 1567 | |
9227dd2a EB |
1568 | if (p->ptrace) |
1569 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
a6f76f23 | 1570 | |
259e5e6c AL |
1571 | /* |
1572 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1573 | * mess up. | |
1574 | */ | |
1d4457f9 | 1575 | if (task_no_new_privs(current)) |
259e5e6c AL |
1576 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1577 | ||
83f62a2e | 1578 | t = p; |
0bf2f3ae | 1579 | n_fs = 1; |
2a4419b5 | 1580 | spin_lock(&p->fs->lock); |
437f7fdb | 1581 | rcu_read_lock(); |
83f62a2e | 1582 | while_each_thread(p, t) { |
0bf2f3ae DH |
1583 | if (t->fs == p->fs) |
1584 | n_fs++; | |
0bf2f3ae | 1585 | } |
437f7fdb | 1586 | rcu_read_unlock(); |
0bf2f3ae | 1587 | |
9e00cdb0 | 1588 | if (p->fs->users > n_fs) |
a6f76f23 | 1589 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1590 | else |
1591 | p->fs->in_exec = 1; | |
2a4419b5 | 1592 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1593 | } |
1594 | ||
56305aa9 | 1595 | static void bprm_fill_uid(struct linux_binprm *bprm, struct file *file) |
8b01fc86 | 1596 | { |
56305aa9 | 1597 | /* Handle suid and sgid on files */ |
1ab29965 | 1598 | struct user_namespace *mnt_userns; |
8b01fc86 JH |
1599 | struct inode *inode; |
1600 | unsigned int mode; | |
1601 | kuid_t uid; | |
1602 | kgid_t gid; | |
1603 | ||
56305aa9 | 1604 | if (!mnt_may_suid(file->f_path.mnt)) |
8b01fc86 JH |
1605 | return; |
1606 | ||
1607 | if (task_no_new_privs(current)) | |
1608 | return; | |
1609 | ||
56305aa9 | 1610 | inode = file->f_path.dentry->d_inode; |
8b01fc86 JH |
1611 | mode = READ_ONCE(inode->i_mode); |
1612 | if (!(mode & (S_ISUID|S_ISGID))) | |
1613 | return; | |
1614 | ||
1ab29965 CB |
1615 | mnt_userns = file_mnt_user_ns(file); |
1616 | ||
8b01fc86 | 1617 | /* Be careful if suid/sgid is set */ |
5955102c | 1618 | inode_lock(inode); |
8b01fc86 JH |
1619 | |
1620 | /* reload atomically mode/uid/gid now that lock held */ | |
1621 | mode = inode->i_mode; | |
1ab29965 CB |
1622 | uid = i_uid_into_mnt(mnt_userns, inode); |
1623 | gid = i_gid_into_mnt(mnt_userns, inode); | |
5955102c | 1624 | inode_unlock(inode); |
8b01fc86 JH |
1625 | |
1626 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1627 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1628 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1629 | return; | |
1630 | ||
1631 | if (mode & S_ISUID) { | |
1632 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1633 | bprm->cred->euid = uid; | |
1634 | } | |
1635 | ||
1636 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1637 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1638 | bprm->cred->egid = gid; | |
1639 | } | |
1640 | } | |
1641 | ||
56305aa9 EB |
1642 | /* |
1643 | * Compute brpm->cred based upon the final binary. | |
1644 | */ | |
1645 | static int bprm_creds_from_file(struct linux_binprm *bprm) | |
1646 | { | |
1647 | /* Compute creds based on which file? */ | |
1648 | struct file *file = bprm->execfd_creds ? bprm->executable : bprm->file; | |
1649 | ||
1650 | bprm_fill_uid(bprm, file); | |
1651 | return security_bprm_creds_from_file(bprm, file); | |
1652 | } | |
1653 | ||
9e00cdb0 ON |
1654 | /* |
1655 | * Fill the binprm structure from the inode. | |
56305aa9 | 1656 | * Read the first BINPRM_BUF_SIZE bytes |
a6f76f23 DH |
1657 | * |
1658 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 | 1659 | */ |
8b72ca90 | 1660 | static int prepare_binprm(struct linux_binprm *bprm) |
1da177e4 | 1661 | { |
bdd1d2d3 | 1662 | loff_t pos = 0; |
1da177e4 | 1663 | |
a6f76f23 | 1664 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
bdd1d2d3 | 1665 | return kernel_read(bprm->file, bprm->buf, BINPRM_BUF_SIZE, &pos); |
1da177e4 LT |
1666 | } |
1667 | ||
4fc75ff4 NP |
1668 | /* |
1669 | * Arguments are '\0' separated strings found at the location bprm->p | |
1670 | * points to; chop off the first by relocating brpm->p to right after | |
1671 | * the first '\0' encountered. | |
1672 | */ | |
b6a2fea3 | 1673 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1674 | { |
b6a2fea3 OW |
1675 | int ret = 0; |
1676 | unsigned long offset; | |
1677 | char *kaddr; | |
1678 | struct page *page; | |
4fc75ff4 | 1679 | |
b6a2fea3 OW |
1680 | if (!bprm->argc) |
1681 | return 0; | |
1da177e4 | 1682 | |
b6a2fea3 OW |
1683 | do { |
1684 | offset = bprm->p & ~PAGE_MASK; | |
1685 | page = get_arg_page(bprm, bprm->p, 0); | |
1686 | if (!page) { | |
1687 | ret = -EFAULT; | |
1688 | goto out; | |
1689 | } | |
e8e3c3d6 | 1690 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1691 | |
b6a2fea3 OW |
1692 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1693 | offset++, bprm->p++) | |
1694 | ; | |
4fc75ff4 | 1695 | |
e8e3c3d6 | 1696 | kunmap_atomic(kaddr); |
b6a2fea3 | 1697 | put_arg_page(page); |
b6a2fea3 | 1698 | } while (offset == PAGE_SIZE); |
4fc75ff4 | 1699 | |
b6a2fea3 OW |
1700 | bprm->p++; |
1701 | bprm->argc--; | |
1702 | ret = 0; | |
4fc75ff4 | 1703 | |
b6a2fea3 OW |
1704 | out: |
1705 | return ret; | |
1da177e4 | 1706 | } |
1da177e4 LT |
1707 | EXPORT_SYMBOL(remove_arg_zero); |
1708 | ||
cb7b6b1c | 1709 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1710 | /* |
1711 | * cycle the list of binary formats handler, until one recognizes the image | |
1712 | */ | |
bc2bf338 | 1713 | static int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1714 | { |
cb7b6b1c | 1715 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1716 | struct linux_binfmt *fmt; |
cb7b6b1c | 1717 | int retval; |
1da177e4 | 1718 | |
8b72ca90 EB |
1719 | retval = prepare_binprm(bprm); |
1720 | if (retval < 0) | |
1721 | return retval; | |
d7402698 | 1722 | |
1da177e4 LT |
1723 | retval = security_bprm_check(bprm); |
1724 | if (retval) | |
1725 | return retval; | |
1726 | ||
1da177e4 | 1727 | retval = -ENOENT; |
cb7b6b1c ON |
1728 | retry: |
1729 | read_lock(&binfmt_lock); | |
1730 | list_for_each_entry(fmt, &formats, lh) { | |
1731 | if (!try_module_get(fmt->module)) | |
1732 | continue; | |
1733 | read_unlock(&binfmt_lock); | |
d53ddd01 | 1734 | |
cb7b6b1c | 1735 | retval = fmt->load_binary(bprm); |
d53ddd01 | 1736 | |
19d860a1 AV |
1737 | read_lock(&binfmt_lock); |
1738 | put_binfmt(fmt); | |
bc2bf338 | 1739 | if (bprm->point_of_no_return || (retval != -ENOEXEC)) { |
19d860a1 | 1740 | read_unlock(&binfmt_lock); |
cb7b6b1c | 1741 | return retval; |
1da177e4 | 1742 | } |
1da177e4 | 1743 | } |
cb7b6b1c ON |
1744 | read_unlock(&binfmt_lock); |
1745 | ||
19d860a1 | 1746 | if (need_retry) { |
cb7b6b1c ON |
1747 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1748 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1749 | return retval; | |
4e0621a0 ON |
1750 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1751 | return retval; | |
cb7b6b1c ON |
1752 | need_retry = false; |
1753 | goto retry; | |
1754 | } | |
1755 | ||
1da177e4 LT |
1756 | return retval; |
1757 | } | |
1da177e4 | 1758 | |
5d1baf3b ON |
1759 | static int exec_binprm(struct linux_binprm *bprm) |
1760 | { | |
1761 | pid_t old_pid, old_vpid; | |
bc2bf338 | 1762 | int ret, depth; |
5d1baf3b ON |
1763 | |
1764 | /* Need to fetch pid before load_binary changes it */ | |
1765 | old_pid = current->pid; | |
1766 | rcu_read_lock(); | |
1767 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1768 | rcu_read_unlock(); | |
1769 | ||
bc2bf338 EB |
1770 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
1771 | for (depth = 0;; depth++) { | |
1772 | struct file *exec; | |
1773 | if (depth > 5) | |
1774 | return -ELOOP; | |
1775 | ||
1776 | ret = search_binary_handler(bprm); | |
1777 | if (ret < 0) | |
1778 | return ret; | |
1779 | if (!bprm->interpreter) | |
1780 | break; | |
1781 | ||
1782 | exec = bprm->file; | |
1783 | bprm->file = bprm->interpreter; | |
1784 | bprm->interpreter = NULL; | |
1785 | ||
1786 | allow_write_access(exec); | |
1787 | if (unlikely(bprm->have_execfd)) { | |
1788 | if (bprm->executable) { | |
1789 | fput(exec); | |
1790 | return -ENOEXEC; | |
1791 | } | |
1792 | bprm->executable = exec; | |
1793 | } else | |
1794 | fput(exec); | |
5d1baf3b ON |
1795 | } |
1796 | ||
bc2bf338 EB |
1797 | audit_bprm(bprm); |
1798 | trace_sched_process_exec(current, old_pid, bprm); | |
1799 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
1800 | proc_exec_connector(current); | |
1801 | return 0; | |
5d1baf3b ON |
1802 | } |
1803 | ||
1da177e4 LT |
1804 | /* |
1805 | * sys_execve() executes a new program. | |
1806 | */ | |
0c9cdff0 EB |
1807 | static int bprm_execve(struct linux_binprm *bprm, |
1808 | int fd, struct filename *filename, int flags) | |
1da177e4 | 1809 | { |
25cf336d | 1810 | struct file *file; |
1da177e4 | 1811 | int retval; |
0f212204 | 1812 | |
a2a8474c ON |
1813 | retval = prepare_bprm_creds(bprm); |
1814 | if (retval) | |
b6043501 | 1815 | return retval; |
498052bb | 1816 | |
9e00cdb0 | 1817 | check_unsafe_exec(bprm); |
a2a8474c | 1818 | current->in_execve = 1; |
a6f76f23 | 1819 | |
25cf336d | 1820 | file = do_open_execat(fd, filename, flags); |
1da177e4 LT |
1821 | retval = PTR_ERR(file); |
1822 | if (IS_ERR(file)) | |
498052bb | 1823 | goto out_unmark; |
1da177e4 LT |
1824 | |
1825 | sched_exec(); | |
1826 | ||
1da177e4 | 1827 | bprm->file = file; |
60d9ad1d EB |
1828 | /* |
1829 | * Record that a name derived from an O_CLOEXEC fd will be | |
b6043501 EB |
1830 | * inaccessible after exec. This allows the code in exec to |
1831 | * choose to fail when the executable is not mmaped into the | |
1832 | * interpreter and an open file descriptor is not passed to | |
1833 | * the interpreter. This makes for a better user experience | |
1834 | * than having the interpreter start and then immediately fail | |
1835 | * when it finds the executable is inaccessible. | |
60d9ad1d | 1836 | */ |
878f12db | 1837 | if (bprm->fdpath && get_close_on_exec(fd)) |
60d9ad1d | 1838 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; |
1da177e4 | 1839 | |
b8bff599 EB |
1840 | /* Set the unchanging part of bprm->cred */ |
1841 | retval = security_bprm_creds_for_exec(bprm); | |
1842 | if (retval) | |
1da177e4 LT |
1843 | goto out; |
1844 | ||
5d1baf3b | 1845 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1846 | if (retval < 0) |
1847 | goto out; | |
1da177e4 | 1848 | |
a6f76f23 | 1849 | /* execve succeeded */ |
498052bb | 1850 | current->fs->in_exec = 0; |
f9ce1f1c | 1851 | current->in_execve = 0; |
d7822b1e | 1852 | rseq_execve(current); |
a6f76f23 | 1853 | acct_update_integrals(current); |
16d51a59 | 1854 | task_numa_free(current, false); |
a6f76f23 | 1855 | return retval; |
1da177e4 | 1856 | |
a6f76f23 | 1857 | out: |
8890b293 | 1858 | /* |
3d742d4b | 1859 | * If past the point of no return ensure the code never |
8890b293 EB |
1860 | * returns to the userspace process. Use an existing fatal |
1861 | * signal if present otherwise terminate the process with | |
1862 | * SIGSEGV. | |
1863 | */ | |
1864 | if (bprm->point_of_no_return && !fatal_signal_pending(current)) | |
4e0608cc | 1865 | force_fatal_sig(SIGSEGV); |
1da177e4 | 1866 | |
498052bb | 1867 | out_unmark: |
9e00cdb0 | 1868 | current->fs->in_exec = 0; |
f9ce1f1c | 1869 | current->in_execve = 0; |
a6f76f23 | 1870 | |
0c9cdff0 EB |
1871 | return retval; |
1872 | } | |
1873 | ||
1874 | static int do_execveat_common(int fd, struct filename *filename, | |
1875 | struct user_arg_ptr argv, | |
1876 | struct user_arg_ptr envp, | |
1877 | int flags) | |
1878 | { | |
1879 | struct linux_binprm *bprm; | |
1880 | int retval; | |
1881 | ||
1882 | if (IS_ERR(filename)) | |
1883 | return PTR_ERR(filename); | |
1884 | ||
1885 | /* | |
1886 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1887 | * set*uid() to execve() because too many poorly written programs | |
1888 | * don't check setuid() return code. Here we additionally recheck | |
1889 | * whether NPROC limit is still exceeded. | |
1890 | */ | |
1891 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
21d1c5e3 | 1892 | is_ucounts_overlimit(current_ucounts(), UCOUNT_RLIMIT_NPROC, rlimit(RLIMIT_NPROC))) { |
0c9cdff0 EB |
1893 | retval = -EAGAIN; |
1894 | goto out_ret; | |
1895 | } | |
1896 | ||
1897 | /* We're below the limit (still or again), so we don't want to make | |
1898 | * further execve() calls fail. */ | |
1899 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1900 | ||
1901 | bprm = alloc_bprm(fd, filename); | |
1902 | if (IS_ERR(bprm)) { | |
1903 | retval = PTR_ERR(bprm); | |
1904 | goto out_ret; | |
1905 | } | |
1906 | ||
d8b9cd54 | 1907 | retval = count(argv, MAX_ARG_STRINGS); |
c07dc0fe KC |
1908 | if (retval == 0) |
1909 | pr_warn_once("process '%s' launched '%s' with NULL argv: empty string added\n", | |
1910 | current->comm, bprm->filename); | |
d8b9cd54 EB |
1911 | if (retval < 0) |
1912 | goto out_free; | |
1913 | bprm->argc = retval; | |
1914 | ||
1915 | retval = count(envp, MAX_ARG_STRINGS); | |
1916 | if (retval < 0) | |
1917 | goto out_free; | |
1918 | bprm->envc = retval; | |
1919 | ||
1920 | retval = bprm_stack_limits(bprm); | |
0c9cdff0 EB |
1921 | if (retval < 0) |
1922 | goto out_free; | |
1923 | ||
1924 | retval = copy_string_kernel(bprm->filename, bprm); | |
1925 | if (retval < 0) | |
1926 | goto out_free; | |
1927 | bprm->exec = bprm->p; | |
1928 | ||
1929 | retval = copy_strings(bprm->envc, envp, bprm); | |
1930 | if (retval < 0) | |
1931 | goto out_free; | |
1932 | ||
1933 | retval = copy_strings(bprm->argc, argv, bprm); | |
1934 | if (retval < 0) | |
1935 | goto out_free; | |
1936 | ||
c07dc0fe KC |
1937 | /* |
1938 | * When argv is empty, add an empty string ("") as argv[0] to | |
1939 | * ensure confused userspace programs that start processing | |
1940 | * from argv[1] won't end up walking envp. See also | |
1941 | * bprm_stack_limits(). | |
1942 | */ | |
1943 | if (bprm->argc == 0) { | |
1944 | retval = copy_string_kernel("", bprm); | |
1945 | if (retval < 0) | |
1946 | goto out_free; | |
1947 | bprm->argc = 1; | |
1948 | } | |
1949 | ||
0c9cdff0 | 1950 | retval = bprm_execve(bprm, fd, filename, flags); |
a6f76f23 | 1951 | out_free: |
08a6fac1 | 1952 | free_bprm(bprm); |
1da177e4 LT |
1953 | |
1954 | out_ret: | |
25cf336d | 1955 | putname(filename); |
1da177e4 LT |
1956 | return retval; |
1957 | } | |
1958 | ||
be619f7f EB |
1959 | int kernel_execve(const char *kernel_filename, |
1960 | const char *const *argv, const char *const *envp) | |
1961 | { | |
1962 | struct filename *filename; | |
1963 | struct linux_binprm *bprm; | |
1964 | int fd = AT_FDCWD; | |
1965 | int retval; | |
1966 | ||
1967 | filename = getname_kernel(kernel_filename); | |
1968 | if (IS_ERR(filename)) | |
1969 | return PTR_ERR(filename); | |
1970 | ||
1971 | bprm = alloc_bprm(fd, filename); | |
1972 | if (IS_ERR(bprm)) { | |
1973 | retval = PTR_ERR(bprm); | |
1974 | goto out_ret; | |
1975 | } | |
1976 | ||
1977 | retval = count_strings_kernel(argv); | |
c07dc0fe KC |
1978 | if (WARN_ON_ONCE(retval == 0)) |
1979 | retval = -EINVAL; | |
be619f7f EB |
1980 | if (retval < 0) |
1981 | goto out_free; | |
1982 | bprm->argc = retval; | |
1983 | ||
1984 | retval = count_strings_kernel(envp); | |
1985 | if (retval < 0) | |
1986 | goto out_free; | |
1987 | bprm->envc = retval; | |
1988 | ||
1989 | retval = bprm_stack_limits(bprm); | |
1990 | if (retval < 0) | |
1991 | goto out_free; | |
1992 | ||
1993 | retval = copy_string_kernel(bprm->filename, bprm); | |
1994 | if (retval < 0) | |
1995 | goto out_free; | |
1996 | bprm->exec = bprm->p; | |
1997 | ||
1998 | retval = copy_strings_kernel(bprm->envc, envp, bprm); | |
1999 | if (retval < 0) | |
2000 | goto out_free; | |
2001 | ||
2002 | retval = copy_strings_kernel(bprm->argc, argv, bprm); | |
2003 | if (retval < 0) | |
2004 | goto out_free; | |
2005 | ||
2006 | retval = bprm_execve(bprm, fd, filename, 0); | |
2007 | out_free: | |
2008 | free_bprm(bprm); | |
2009 | out_ret: | |
2010 | putname(filename); | |
2011 | return retval; | |
2012 | } | |
2013 | ||
2014 | static int do_execve(struct filename *filename, | |
ba2d0162 | 2015 | const char __user *const __user *__argv, |
da3d4c5f | 2016 | const char __user *const __user *__envp) |
ba2d0162 | 2017 | { |
0e028465 ON |
2018 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
2019 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
2020 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
2021 | } | |
2022 | ||
be619f7f | 2023 | static int do_execveat(int fd, struct filename *filename, |
51f39a1f DD |
2024 | const char __user *const __user *__argv, |
2025 | const char __user *const __user *__envp, | |
2026 | int flags) | |
2027 | { | |
2028 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
2029 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
2030 | ||
2031 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
2032 | } |
2033 | ||
2034 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 2035 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 2036 | const compat_uptr_t __user *__argv, |
d03d26e5 | 2037 | const compat_uptr_t __user *__envp) |
0e028465 ON |
2038 | { |
2039 | struct user_arg_ptr argv = { | |
2040 | .is_compat = true, | |
2041 | .ptr.compat = __argv, | |
2042 | }; | |
2043 | struct user_arg_ptr envp = { | |
2044 | .is_compat = true, | |
2045 | .ptr.compat = __envp, | |
2046 | }; | |
51f39a1f DD |
2047 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
2048 | } | |
2049 | ||
2050 | static int compat_do_execveat(int fd, struct filename *filename, | |
2051 | const compat_uptr_t __user *__argv, | |
2052 | const compat_uptr_t __user *__envp, | |
2053 | int flags) | |
2054 | { | |
2055 | struct user_arg_ptr argv = { | |
2056 | .is_compat = true, | |
2057 | .ptr.compat = __argv, | |
2058 | }; | |
2059 | struct user_arg_ptr envp = { | |
2060 | .is_compat = true, | |
2061 | .ptr.compat = __envp, | |
2062 | }; | |
2063 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 2064 | } |
0e028465 | 2065 | #endif |
ba2d0162 | 2066 | |
964ee7df | 2067 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 2068 | { |
801460d0 HS |
2069 | struct mm_struct *mm = current->mm; |
2070 | ||
2071 | if (mm->binfmt) | |
2072 | module_put(mm->binfmt->module); | |
1da177e4 | 2073 | |
801460d0 | 2074 | mm->binfmt = new; |
964ee7df ON |
2075 | if (new) |
2076 | __module_get(new->module); | |
1da177e4 | 2077 | } |
1da177e4 LT |
2078 | EXPORT_SYMBOL(set_binfmt); |
2079 | ||
6c5d5238 | 2080 | /* |
7288e118 | 2081 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
2082 | */ |
2083 | void set_dumpable(struct mm_struct *mm, int value) | |
2084 | { | |
7288e118 ON |
2085 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
2086 | return; | |
2087 | ||
26e15225 | 2088 | set_mask_bits(&mm->flags, MMF_DUMPABLE_MASK, value); |
6c5d5238 | 2089 | } |
6c5d5238 | 2090 | |
38b983b3 AV |
2091 | SYSCALL_DEFINE3(execve, |
2092 | const char __user *, filename, | |
2093 | const char __user *const __user *, argv, | |
2094 | const char __user *const __user *, envp) | |
2095 | { | |
c4ad8f98 | 2096 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 2097 | } |
51f39a1f DD |
2098 | |
2099 | SYSCALL_DEFINE5(execveat, | |
2100 | int, fd, const char __user *, filename, | |
2101 | const char __user *const __user *, argv, | |
2102 | const char __user *const __user *, envp, | |
2103 | int, flags) | |
2104 | { | |
51f39a1f | 2105 | return do_execveat(fd, |
8228e2c3 | 2106 | getname_uflags(filename, flags), |
51f39a1f DD |
2107 | argv, envp, flags); |
2108 | } | |
2109 | ||
38b983b3 | 2110 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
2111 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
2112 | const compat_uptr_t __user *, argv, | |
2113 | const compat_uptr_t __user *, envp) | |
38b983b3 | 2114 | { |
c4ad8f98 | 2115 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 2116 | } |
51f39a1f DD |
2117 | |
2118 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
2119 | const char __user *, filename, | |
2120 | const compat_uptr_t __user *, argv, | |
2121 | const compat_uptr_t __user *, envp, | |
2122 | int, flags) | |
2123 | { | |
51f39a1f | 2124 | return compat_do_execveat(fd, |
8228e2c3 | 2125 | getname_uflags(filename, flags), |
51f39a1f DD |
2126 | argv, envp, flags); |
2127 | } | |
38b983b3 | 2128 | #endif |